Towards Automated Embedded Systems Programming

Abstract

Writing code for embedded systems poses unique challenges due to hardware involvement. Developers often need to learn domain-specific knowledge to write embedded codes. Learning such knowledge is time-consuming and hinders de-velopers' productivity. This paper presents a proposal for an automated code generation approach, specifically designed for embedded systems. The work is composed of three milestones, i.e., understanding the needs of embedded developers by analyzing posts from discussion forums, developing a tool to recommend driver libraries of 110 hardware and generate its interface configurations and usage patterns, and improving the generation accuracy of the prior tool using program analysis techniques. The tool will be evaluated using various metrics from machine translation.. classification.. and information retrieval fields.